Professional Documents
Culture Documents
Search
Collections
Journals
About
Contact us
My IOPscience
A Numerical Procedure for Flow Distribution and Pressure Drops for U and Z Type
Configurations Plate Heat Exchangers with Variable Coefficients
This content has been downloaded from IOPscience. Please scroll down to see the full text.
2012 J. Phys.: Conf. Ser. 395 012060
(http://iopscience.iop.org/1742-6596/395/1/012060)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 58.68.86.226
This content was downloaded on 26/02/2014 at 07:13
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
1. Introduction
Manifold flow distributed systems are extensively used in a wide range of engineering applications.
Among them we can mention: Plate heat exchangers, micro channel heat exchangers, fuel cells, solar
collectors, etc. In these devices, where two manifolds are interconnected with several channels,it is
important to determine the flow distribution and pressure drops through each channel between both
manifolds, because they affect the efficiency and performance of the system [1] and are a basic
framework for further thermal analysis.
The problem of exactly determining these magnitudes is a difficult task since the flow is 3dimensional, of elliptic character and, in some cases, can be transient depending on the operating
conditions. The system has to be modelled in such a way that the information obtained represents with
accuracy the real flow in a reasonable computing time. In the past fifty years the problem has been
studied using three different approaches [4];(i) computational fluid dynamics (CFD), (ii) analytical
models and (iii) discrete models. The CFD approach consists in solving 2-D or 3-D momentum,
energy, mass and state equations, with turbulence models for closing the problem. This model
provides the most detailed flow and offers the advantage of not requiring knowing in advance the loss
associated to the flow turning between manifold and channel neither the overall momentum nor
friction coefficients that consider the irreversibilities associated with separated flow. However, the
computational cost and empiricism of the turbulence models are limiting factors of this kind of
approach.
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
On the other hand, analytical and discrete approaches rely on 1-D formulations considering mean
values of velocities and pressure at each section, taking advantage of the slenderness of the manifold
and channel tubing. Obviously, the flow is far away from being 1-D, so it is necessary to introduce
well founded coefficients to ensure that the 1-D flow approaches the real flow. The analytical models
are particular cases of the governing equations of some discrete models. These analytical models give
the designer a simplified view on how the fluid is distributed among channels and to estimate the
pressure losses, but no analytical solution has been found for the case when the coefficients are
correlated with the flow variables. In this scenario, discrete models incorporate the capability to solve
problems with variable coefficients of any kind. This, jointly with the development of numerical
algorithms and the nowadays computational power available make this kind of approach a very
promising tool for analyzing and optimizing manifold systems.
Another factor influencing the velocity and pressure distributions is the arrangement type; the two
basic ones are the U and Z type. The difference between them is how the fluid exits the manifold;
as Figure 1 shows.
Z
a)
b)
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
the flow distribution system is considered as an infinite succession of channels in a continuum fashion.
This over-predicts the pressure drop, due to the lateral turning loss coefficient (Cd). More recently,
Wang [4] developed a numerical procedure with the analytical solution for friction in the manifold
[12,13] using a predictor-corrector strategy to obtain distributions allowing variable friction factors
and pressure change coefficient (), see Eq. 1. In addition, Bajura and Jones [14] first proposed models
considering discrete channels for both combining and dividing lateral inlets. They also offered an
extensive compilation of experimental work, introducing the pressure change coefficient in the
models, wich take into account the non-conservation of axial linear momentum in the manifolds.
Subsequently, they developed a model for U and Z type configurations [2]. They also performed an
experimental research on combining and dividing manifold systems. One of the most useful results in
this work was the measurement of the overall momentum coefficient (CT) in Eq. 1, which quantifies
the axial linear momentum transported through the channel, establishing this parameter as
1.050.05and 2.660.05, for dividing and combining manifold respectively. In a further study, Datta
and Majumdar [3] proposed a discrete model via the finite differences method based on the nonvariable overall momentum coefficient version of the Bajura and Jones model [2]. Their results show
good agreement with experimental data.
3. Present work
The present work proposes a 1-D, steady-state and parabolic discrete model introducing the novelty of
allowing for variable overall momentum coefficient and friction factors in the manifold. The
discretization is made via finite differences method. The coupling of velocities and pressure is
implemented with a pressure correction (SIMPLE) algorithm, Patankar [15]. Another feature is that
the equations and boundary conditions are expressed with the use of the auxiliary functions , and .
They allow a compact formulation for Z and U distribution system types.
3.1. Overall momentum coefficient and friction factor
Inside a manifold, the flow is not uniform in every section, as assumed by the 1-D simplification. In
fact, it has a certain transversal profile. The lower velocity near the walls due to viscous effects
implies less energy, so when the fluid passes a channel inlet, the fluid with less axial velocity will
enter the channel. That entrance is not instantly parallel to the channel axis because of inertia, so part
of the upstream axial energy is transferred to the mass that enters the channels. In addition, an error
can be made in quantifying the upstream axial momentum because of the uniform velocity profile
simplification. So in order to correct this, a coefficient is introduced for the header flow; the so-called
header momentum coefficient () or simply the Boussinesq coefficient. Mathematically, the header
momentum, pressure change () and overall momentum coefficient (CT) are defined as:
1
1
= 2 W 2 dA =
Wc ( Ac ) U c dAc
W A Ac
W U c A c Ac
1
CT = ( 2 )
(1)
2
It is important to mention that =0 implies that the manifold flow reaches the maximum static pressure
recovery downstream the channel inlet.
By observations made by Bajura and Jones [2], the overall momentum coefficient CT is approximately
fixed when the channel/manifold diameter ratio is greater than 0.5. Those values in CT are 1.05 and
2.66 for dividing and combining manifolds respectively, but in order to explore a possible
enhancement, in the present work we used a linear variation of CT from 1.3 to 1.05 between z = 0 and z
= 0.2 to take into account the readjustment of the velocity profile after the first few laterals channel
inlets [14]. The values of d/dz are obtained differentiating Eq. 1 assuming that is constant and equal
to 0.8 [6,7]. In the combining manifold, the value of CT is fixed to 2.66 [14] and the values of are
calculated from the experimental values given by McNown [7] and Zeisser [6]. It is relevant to notice
that the present formulation is not restricted to a linear distribution of CT; and other kind of functions
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
can be introduced. The friction factors in the manifolds and the channels have been determined with
the Blasius formula for the manifold and the Rao and Sunden [1] correlation for the channels:
(2)
f c = 21.41 Rec 0.301
f = 0.3164 Re0.25
(1)
Where Rec is the Channel Reynolds number based on twice the plate spacing.
4. Mathematical formulation
For the mathematical formulation here proposed, steady-state, 1-D and incompressible flow has benn
assumed. The analysis uses the control volume shown in Fig. 2 where, the total mass enters through
section A, some fluid goes through section B and the remaining enters the channel through section C
with a certain angle implying components of velocity Wc and Uc. It has been considered that the shear
stress at the walls can be approximated by the Darcy-Weisbach formulation (w=fW2/8). With these
assumptions, applying conservation of mass and linear momentum, the governing equations in
dimensionless form are:
d CT w2
dC
dp L f S d 2
w2 T =
+
(3)
w
dz
dz
dz 8 A
dz
uc Ac
dw
=
(4)
z A
dz
Equation 4 is a modified version of the Bajura and Jones model, for further details about the derivation
of the original version of Eqs. 4 and 5, see Bajura and Jones [2]. The modification consists in
introducing inside the convective term the overall momentum coefficient (CT). To do so, the chain rule
is applied to the original convective term. The introduction of this factor inside the convective term
allows an easy integration over the control volume when it is variable.
Z
P(z)
W
INLET
W(z)
W(z+z)
Z
P(z+z)
INLET
C.V.
P(z)
W(z)
Wc
Ac
Uc
Uc
Wc
C.V.
Wc
P*(z)
W*(z+z)
C.V.
Wc
Uc
OUTLET
Uc
Ac
W*(z)
P(z+z)
C.V.
Uc
A*
W(z+z)
Wc
W0*
P*(z)
P*(z+z)
A*
Uc
W*(z)
W*(z+z)
P*(z+z)
a)
W0*
OUTLET
C.V.
b)
Figure 2. Combining and dividing flows control volumes for: a) Type U, b) Type Z
4.1. Boundary conditions
The velocities can be specified at both boundaries, for either a dividing or a combining manifold. In
particular, the pressure can be specified only in z=0 for the dividing manifold and in z=1 for the
combining manifold for type U or z=0 for type Z configuration in the combining manifold. The
generalized boundary conditions for velocity and pressure in dimensionless form are:
1
1
1
A
A
w ( z = 0, , ) = (1 + ) + (1 )
w ( z = 1, , ) = (1 + ) (1 )
2
2
4
A*
A*
p ( z = 0, dividing ) = 0
p* ( z = 0, combining ) = 0 if Z type
*
p ( z = 1, combining ) = 0 if U type
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
= ( 1)
= ( 1)
2 if Dividing Manifold
2 if U Type
=
1 if Combining Manifold
1 if Z Type
2 if Dividing Manifold
1 if Combining Manifold
(5)
( uc )i = Cd {2 ( pi pi* )}
1/2
(6)
f L
Where Cd is the channel discharge coefficient, defined as: Cd = c c +
d c i
j
j =1
1/2
Aw = ( C T
) P A w wk 1
AEe = ( C T
L f S d k 1
Sc =
+
we
dz
8A
Pp
We
Ae = Aw + AEe + S p
L f S d k 1
Sp =
+
we
dz
4A
Uci+2
Uci+1
Wci+1
Uci
Wci
Ww
) E A wek 1
Uci
Wci
Wci+2
W Ee
PE
Ww
We
Pp
a)
b)
Where
De =
AEe
Ae
Dw =
Aw
Ae
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
pP = p 'P + pP
(9)
6. Numerical experiments
This section offers a comparison of the present model with experimental and analytical results. There
is lack of experimental data for channel velocity and pressure distributions in plate heat exchangers.
These devices can be seen as a manifold system with high lateral resistance, since; the friction in the
channels is in the order of 50 times the friction in the manifolds. Because of that, the first case of our
study consists in the numerical replication of the high lateral resistance experiment made by Bajura
and Jones [14]. For this numerical experiment we present the dimensionless pressure distributions (Pc,
Pd and P, see Eq. 11). The comparison is also made with the numerical results with constant
coefficients of Datta and Majumdar [3]. Later on, the dimensionless channel velocity (uc) distribution
is compared with the analytical solution of Bassiouny and Martin [10, 11], obtaining the characteristic
flow parameter m2 as the solutions of Eq. (12) or (13). The characteristics of this case are given in
Table 1.
Table 1. Characteristics for the high lateral resistance experiment (case 1).
dc D
l D
Cd
Exp. H
-1
8.98 ms
0.106 m 1.56 m
0.375
2.55
12.2
0.4048
The second and third case of study in this paper consist in the numerical replication of the experiments
made by Rao and Sunden [1], in the second case the influence of the number of channels and mean
channel Reynolds number is studied. Finally, the third case is the study of the dimensionless channel
velocity (in respect to the mean channel velocity) distribution for port diameters of 70 mm and 35 mm
and inflow conditions of 3.6 Ls-1 and 0.13 Ls-1. The characteristics for these two cases are presented
in Table 2. In all cases the study was made for both U and Z type configurations.
Table 2. Geometric characteristics of the plate for cases 2 and 3.
70 - 35 mm
Port diameter
0.6 m
Vertical distance between ports
60
Chevron angle
2.9 mm
Spacing between plates
Hydraulic diameter of the channels
5.8 mm
0.4 mm
Plate width
Number of plates (N)
10, 15 or 18
p * pexit
p pexit
p p*
pc =
pd =
p =
(10)
pinlet pexit
pinlet pexit
max [ p p *]
Where:
A
m
pinlet p *exit =
for U type
2 N Ac tanh ( m )
2
2
2
c A m
1 sec h ( m )
=
+ 1 + 2
2 N Ac tanh ( m )
m tanh ( m )
pinlet p *exit
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
2 Ac N
c A
(11)
for Z type
(12)
Pc, Pd, P
7. Discussion of results
Figs. 4.a and 4.b show the dimensionless pressure for U and Z type for case 1. For both of them, the
predictions are in good agreement with experimental and numerical data, but, in figure 4b the
numerical results with constant coefficients (Datta and Majumdar) in the dividing manifold, pressure
distribution are slightly more in accordance with experimental data. One reason for this discrepancy is
the linear overall distribution function assumed in this work, the real form of this function is of
different shape and therefore must be established empirically. Fig. 5 shows the dimensionless channel
velocity distribution, calculated with the present model, and the analytical solution of the Bassiouny
and Martin model. The analytical velocity profiles are calculated with the value of the flow parameter
m2 determined with the numerical pressure distribution (Figs. 4.a and 4.b). In these figures a greater
deviation in the Z type configuration is observed. The deviation can be in part because the analytical
model does not take into account the friction losses in the manifolds. However, the behaviour of the
present numerical model corresponds to the one predicted by the analytical one.
1.6
PRESENT WORK
PRESENT WORK
Pd
DATTA & MAJUMDAR
DATTA & MAJUMDAR
Pd
1.2
EXPERIMENTAL
Cd=0.4155
Cd=0.4155
Pc, Pd, P
EXPERIMENTAL
1.2
Pc
0.8
0.8
0.4
0.4
Pc
0
0
2
10
12
14
16
18
a)
20
10
12
14
16
18
b)
20
0.6
Uc
Uc
Present Work
Bassiouny and Martin
Present Work
Bassiouny and Martin
0.45
0.4
0.3
0.2
4
12
a)
16
0.15
20
12
16
b)
20
Figure 5. Channel velocity distribution comparison with analytical models a) U type, b) Z type.
Fig.6 presents the total pressure drop in the modelled heat exchanger and again the deviation from
experimental results is larger for the Z type configuration but it is also shown that this difference is not
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
50
70
p0-p0*
p0-p0*
only affected by the configuration type, but also by the number of channels and the channel Reynolds
number (as a consequence of the inflow Reynolds number). Finally, Figs. 7 and 8, show that for port
diameters of 70 mm the model reproduces almost perfectly the experiments but for diameters of 35
mm, predictions of the channel velocity distribution slightly differ from the experimental
measurements. This occurs because in the calculation of the lateral discharge coefficient, the pressure
loss due to sudden expansion or contraction in the channels was neglected; and for lower port diameter
this expansion or contraction is greater, so the velocity in the first channels and the last ones for U and
Z type configurations respectively are greater than the prediction made. From the results obtained it
can be said that the present model reproduces with a good degree of accuracy the experiments.
Moreover a model with variable coefficients has been obtained, but in order to improve the model it is
necessary to find either analytically or experimentally the form of the overall momentum coefficient
distribution along the dividing manifold.
Present Work
Exp. Rao & Sunden 2006
60
10 Channels
Present Work
Exp. Rao & Sunden 2006
40
10 Channels
50
40
30
15 Channels
30
15 Channels
20
20
18 Channels
10
10
1000
2000
3000
a)
4000
Rec
5000
1000
2000
3000
b)
4000
Rec
5000
Figure 6. Dimensionless total pressure drops with a port diameter of 70 mm: a) U type, b) Z type.
1.2
Uc
1.2
Uc
Q=0.13 L/s
1.1
Q=3.6 L/s
0.9
0.9
0
a)
10
15
b)
10
15
Figure 7. Dimensionless channel velocity distribution for Type U with 15 channels: a) Q=3.6 Ls-1, b) Q=0.13 Ls-1
1.3
Uc
1.3
1.2
Uc
1.2
1.1
0.9
0.9
Q=0.13 L/s
Q=3.6 L/s
0.8
0.8
0
a)
10
15
b)
10
15
Figure 8. Dimensionless channel velocity distribution for Type Z with 15 channels: a) Q=3.6 Ls-1, b) Q=0.13 Ls-1
8. Conclusions
On the grounds of the detailed comparison shown, it can be concluded that with the model developed:
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
a) The predictions are in accordance with experimental data and are especially accurate for U
type configurations.
b) The velocity distributions are in accordance with the velocity predicted by the Bassiouny
models.
c) To further improve the model it is desirable to find an analytical or experimental model for the
overall momentum coefficient distribution along the dividing manifold.
References
[1] Rao B, Sunden B, Das S 2005 An experimental and theoretical Investigation of the effect of flow
maldistribution on the thermal performance of Plate Heat Exchangers J. Heat Transfer, 127(3), 33245.
[2] Bajura R A, Jones E H 1976 Flow distribution manifolds J. Fluids Eng. 98(4) 654-665.
[3] Datta A B, Majumdar A K 1980 Flow distribution in parallel and reverse flow manifolds Int. J.
Heat & Fluid Flow 2(4) 253-262.
[4] Wang J 2011 Theory of flow distribution manifolds Chemical engineering science 168 1331-45.
[5] Kubo T, Ueda T 1969 On the characteristics of divided flow and confluent flow in headers Bulletin
of JSME, 12(52) 802-809.
[6] Zeisser M H Summary Report of Single-Tube Branch and Multi-Tube Branch Water Flow Tests
conducted by the University of Connecticut Pratt and Whitney aircraft division, United aircraft
corporation, Report No PWAC-231 USAEC Contract AT(11-1)-229, May 1963.
[7] McNown J S 1954 Mechanics of manifold flow Transactions ASCE 119 1103-42.
[8] Markland E 1959 Analysis of flow from pipe manifolds Engineering 187 150-151.
[9] Acrivos A, Babcock B D and Pigford R L 1959 Flow Distribution in manifolds Chemical
engineering science 10 112-124.
[10] Bassiouny M K, Martin H, 1984 Flow distribution and pressure drop in plate heat exchangers-I: U
type arrangement Chemical engineering science 39(4) 693-700.
[11] Bassiouny M K, Martin H, 1984 Flow distribution and pressure drop in plate heat exchangers-II:
Z type arrangement Chemical engineering science 39(4) 701-704.
[12] Wang J, 2008 Pressure drop and flow distribution in parallel-channel fuel cell stacks: U-type
arrangement. Int. J. of Hydrogen Energy, 33(21) 6339-6350.
[13] Wang J, 2010 Pressure drop and flow distribution in parallel-channel configurations of fuel cells:
Z-type arrangement. Int. J. of Hydrogen Energy, 35(11) 5498-5509
[14] Bajura R A, Jones E H 1971 A model for flow distribution in manifolds Journal of Engineering
for Power Fluids Eng Trans ASME 93(1) 7-12.
[15] Patankar S 1980 Numerical Heat Transfer and Fluid Flow McGraw Hill.
Nomenclature
A
A,D,S
Re Reynolds number
/
S
Perimeter
CT
Uc
Cd
uc
dc
Friction factor
Subscript
Channel, combining
manifold
Dividing manifold
Dimensionless channel
velocity (Uc/W0)
i, j
Index
W, E, P
(=Uc,i/N)
w Dimensionless Axial
velocity (=W/W0)
w,e,Ee
IOP Publishing
doi:10.1088/1742-6596/395/1/012060
Axial coordinate ( m )
Length of manifold
Dimensionless axial
coordinate (=Z/L)
Superscripts
Lc
Length of channel
m2
Number of channels
Dimensionless pressure
/
Greek symbols
Header momentum
coefficient
Pressure change
coefficient.
combining and dividing
functions
Finite difference
10
Inlet conditions
(*)
()
Mean value
( ')
Correction
Arrangement
functions